Sources of variability in consonant perception of normal-hearing listeners

Abstract

Responses obtained in consonant perception experiments typically show a large variability across stimuli of the same phonetic identity. The present study investigated the influence of different potential sources of this response variability. It was distinguished between source-induced variability, referring to perceptual differences caused by acoustical differences in the speech tokens and/or the masking noise tokens, and receiver-related variability, referring to perceptual differences caused by within- and across-listener uncertainty. Consonant-vowel combinations consisting of 15 consonants followed by the vowel /i/ were spoken by two talkers and presented to eight normal-hearing listeners both in quiet and in white noise at six different signal-to-noise ratios. The obtained responses were analyzed with respect to the different sources of variability using a measure of the perceptual distance between responses. The speech-induced variability across and within talkers and the across-listener variability were substantial and of similar magnitude. The noise-induced variability, obtained with time-shifted realizations of the same random process, was smaller but significantly larger than the amount of within-listener variability, which represented the smallest effect. The results have implications for the design of consonant perception experiments and provide constraints for future models of consonant perception.

Vowel and consonant confusions from spectrally manipulated stimuli designed to simulate poor cochlear implant electrode-neuron interfaces

“…This analysis compares confusion matrices resulting from distinct conditions of spectral manipulation to evaluate the difference in phoneme perception between the two conditions (e.g., Zaar and Dau, 2015). Each cell of one confusion matrix is compared to the corresponding cell in the other confusion matrix to determine the overall difference in phoneme perception between the two matrices, scaled from 0% to 100%: a distance of 0% indicates that the matrices compared are exactly the same, and a distance of 100% indicates complete dissimilarity between the matrices.…”

“…Baseline perceptual distance for the analyzed matrices is determined by calculating the within-subject perceptual distance. For example, Zaar and Dau (2015) applied this process to consonant identification confusion matrices to determine the effects of different talkers, types of noise used, and listeners on consonant perception. They calculated the perceptual distance within each subject from test and retest runs and used the resulting values as a baseline for other measurements, since these values represent listener uncertainty.…”

Speech recognition error patterns for steady-state noise and interrupted speech

“…While very detailed error analyses have been conducted for the perception of nonsense syllables (e.g., Miller and Nicely, 1955;Phatak and Allen, 2007;Zaar and Dau, 2015), error analysis for the perception of sentencelength materials has largely relied on binary classification of responses. Importantly, however, examining perceptual errors can further inform our understanding of how listeners resolve speech across different listening environments.…”

A Danish Nonsense Word Corpus for Phoneme Recognition Measurements

“…Large variations in the recognisability of different consonants have previously been demonstrated, e.g. [4,20,21]. In the present study, recognition scores for the individual target phonemes in the GC word lists of the four talkers (averaged across listeners) were determined and an SRT was estimated for each phoneme.…”

“…A listener may misperceive a word if a single phoneme is incorrectly identified; the human auditory system is thus naturally tuned to the identification of these. Rather than the 'macroscopic' view, as for example reflected in sentence intelligibility tests, analysing phoneme perception can be considered a 'microscopic' view of intelligibility [4]. A speech-processing algorithm aimed at improving the perception of highfrequency phonemes should ideally achieve this without affecting the perception of other phonemes.…”